Large Language Models Bootcamp Information Session

Key Takeaways

all right perfect we have people joining in the zoom chat and we are going to go live on our social media in just a minute thank you Ramin and Ramin if you can let me know once we are live that would make me make it easier for me yes Roger we live uh you're on mute okay now okay uh so so we'll go ahead and get started uh welcome to the webinar everyone my name is rajal I am one of the lead instructors at data science Dojo um what we are going to do today is uh we'll go over some of the large language model uh offerings that we have um uh the offerings actually include uh our inperson boot camp uh a 5-day 40-hour boot camp and in addition to that we have some online offerings that are there um consider them to be shorter versions of the the same boot camp uh I will go over those one by one in a moment so um I I think if you are joining here most likely since we don't run any ads most likely you already know who we are but for those of you who may not be aware we have been around for quite some time uh we are one of the uh oldest and one of the most recognized company in the uh in this uh boot camp and uh data science and machine learning and now large language models boot camp space as a matter of fact we are the only boot camp uh in in the world right now that uh offers a large language models boot camp um so we have 11,000 plus graduates uh and um more than 3,000 companies uh that we have trained globally and uh pretty much uh any country on the planet it has been someone from the any of the every country has repres has been represented in one of our train ings um so where did this boot camp uh actually how did this the Genesis of the boot camp so the um the way this boot camp started was actually um due to the fact that uh the our the services side of our business it has been helping a lot of uh um the service side of our business had been helping a lot of uh customers uh Enterprise customers build um their llm applications and as a result of that we were um just a minute Ramin can you confirm that you can see my screen yeah yes Raja your screen is visible okay that sounds good okay so so we have been building these Solutions um uh for our Enterprise customers um and what we realized is that uh you know while Bard and chat GPT and anthropic they make uh they make uh make the task of building these Solutions or building an a chatbot uh quite easy but when you go to Enterprise setting there are there are a lot of challenges uh let's talk about the regulatory challenges let's talk about you know the IP and proprietary data type challenges there are a lot of challenges that uh and data governance for that matter right so um there are a lot of challenges that go beyond Bas just that basic chat GPT or B like experience um there are a lot of practical constraints uh you may or may not realize uh but the token consumption and the and the usage and the cost the these tools are incredibly expensive and when you start adopting these tools at Enterprise scale there are a lot of architectural consider considerations that you have to worry about then um coming up with the right prompts right prompt structure um how do you evaluate how do you create your evaluation data sets whether you use an open source or close Source model um how do you control uh hallucination how do you keep the knowledge updated um how do you find tuna model when to find tuna model when to use retrieval augmented generation uh how to uh whether to use only Vector store or use uh keyword uh based or lexical search store or use a h hybrid search uh so all those considerations for any Enterprise application they are actually uh they are actually very difficult trade-offs to make and what we do in this boot camp the the 40-hour version of the training is that we cover all of these topics end to end uh so I will go over this uh these topics uh I'll give you a very high level idea of the culum and then uh happy to take any questions and meanwhile if there are any questions I will keep uh looking at my Q&A here uh on those who are joining by Zoom I will keep looking at the Q&A but for those of you who are on uh on joining through any of the social media channels please type in your questions there and the team is going to Route the questions to me so um so let's start with our large language models boot camp um the the boot camp itself uh is um as I said there is the inperson experience and there is an instructor LED online experience so first I will talk about the general curriculum uh regardless and then we'll start with uh you know going you know what is the online version what is the in-person version so the in the in the in-person uh version of the boot camp we have it's a 40-hour training you come in uh and you start your day on a Monday at at 8:00 a.m. and you finish your day on uh on a Friday at 5:00 p.m. um we uh we take care of everything for you right so you know we have uh this uh uh wherever you come in you know we take care of pretty much like everything your your your your food any software uh you know any coffee beverages uh you know sandboxes any tools that you need any cloud computing costs pretty much everything that you would need it's basically you just need a browser enabled computer and that's it you just bring your laptop and uh that's it U so it's a comprehensive very comprehensive curriculum uh we have designed the curriculum we are constantly reviewing the curriculum with the help of our partners that you will see shortly uh it is uh I have to say it is the only curriculum so you know there's not much competition out there but despite that I actually sincerely feel I mean honestly feel that this is one of the most comprehensive curriculums anything that that we think and we bump into while we build solutions for our customers we actually come back and incorporate that in our curriculum and I will show you uh what I mean shortly uh so for four days you're going to um learn uh Implement learn Implement learn Implement and on the fifth day you are going to be building uh an actual llm application um and then basically a working prototype of an LM llm application um as I said uh when you register for the boot camp you are elig eligible for $500 uh credit for all software and Cloud servicing cloud services needs um U thanks to our partners so you can see some of the partners I don't think I think I don't think all the partners are here but I mean we have a few Partners who are uh we are discussing and we should see more partners and I will mention to you how these Partners come into play what are they doing uh and you know and for instance why rpod what do they do right so why mongodb I mean how are they related to large language model space and so on so um um the the boot camp uh the the fees uh the registration fee covers uh uh all the token consumptions that you would uh all the token consumption cost all the GPU uh Cloud usage um and all the infrastructure and plus as I said I mean hosting you uh your uh your uh you know your breakfast lunch uh dinner your um you know all everything that you would need during that week um we also actually take pride in the fact that we have some of the industry leaders who would come in and they will uh give talks from real case studies uh from big companies uh we have people coming from uh stripe the payment processing company mongodb we have had guest talks from there we have symol ai they have had guest talks we have had guest talks from there vecta came in for a guest talk and I can keep going so but the idea is that we we we have this firm conven conviction that uh if you want to learn uh a real world application is much more than just technology you have to understand how to how uh how software or how llm application is built in real life so for to that end we bring in people real case studies from big companies um uh we will talk about all the you know all the nuances of not just the engineering issues but also the cultural uh challenges and adoption Etc uh Hands-On exercises are there for all topics wherever it is relevant I will show you our infrastructure and uh as I said earlier as well you will be building an llm app application at the end of the boot camp and I think one slide that I we should have added is that you will be getting uh once you finish uh the training you will be getting a uh verified certificate from the University of New Mexico continuing education they are our academic Partners so it's once again part of the program you register you finish the training you get a certificate uh and the certificate is recorded in their registar office you can request a transcript at the end of the uh once it is reported of course a few days or weeks after the training I don't know exactly how many but I mean maybe one to two weeks after the training uh let me go ahead and get started right so what I'm going to do here is uh uh I will give you an idea right so of the ecosystem uh this entire llm application ecosystem so the at the core of this entire uh ecosystem the architecture um we have these things what we call the llm uh large language models I mean call them the foundation models and uh openi has their Foundation model Google has their Foundation model anthropic is there uh then uh there are other proprietary models uh uh there is open-source models like mixol um llama llama 2 Series uh so U but these models alone are not enough to build an application what you need to have is you need to have a vector database along with it and then this ecosystem is once again a huge giant ecosystem uh even some of the some of the traditional uh relational database players or nosql database players they are also venturing into the vector database market for instance you saw mongodb um uh being our partner uh for the delivery of the boot camp because also mongodb has also added the vector functionality um then postgress has added uh Vector functionality pretty much all the mainstream databases are actually adding functionalities functionality for uh Vector uh Vector search and um and we get into the vector database business how you index uh how you retrieve how you uh store vectors and all the nuances uh along with also uh semantic caching we also talk about in detail what is semantic caching and how do you how would you actually builda semantic cache right um so we talk about vectors uh we talk about in detail about embeddings some models are both encoder decoder models some are uh encoder only models um so we will get into the business of uh uh embeddings uh in detail um and then we have different instructors we are actually very excited uh we have L Sano from coare he is one of the best people uh to teach um uh embeddings and then he's the one who actually leads the session on embeddings then on Vector databases you know it is a few different people we have in the past I mean I've taught it Sanjay has taught it then Sam party he has taught this and Sam party from reddis uh so we have different people who have been uh teaching different components uh now there is this component of logging and llm Ops uh or and guardrails how do you actually uh how do you make sure that when a prompt is given uh a prompt that is unreasonable a prompt that does not comply with your prompt policy right so let's say you have an application internally deployed uh in your Enterprise uh you do not want people to query certain things maybe uh don't no one can ask for a phone number no one can ask for any Politically Incorrect thing so uh the idea of um putting guardrails around uh preventing uh any toxic prompts any toxic responses how do you actually uh measure token consumption how do you log uh how do you deploy a model all of that um so we'll we'll talk we talk about all of that in addition to that uh we talk about uh uh Frameworks that are actually that make building uh may make building rag applications easier uh things like Lang chain we actually spent quite a bit of time on Lang chain we uh cover uh Lang chain in a lot of depth um so basically the I cannot I can spend maybe 30 minutes just on this Slide the idea is that we are not just you know a basic prompt engineering course we are we are in the business of teaching people enabling people to teach them uh pretty much the entire bread breadth or entire spectrum of possible tools and Technologies uh and give them give everyone some practical experience so they can be on their own uh uh in building the applications uh after the training so let me go over and give you an overview of the curriculum I'm going to instead of showing you this on a slide I'm going to actually let me actually maybe I can Flash some of the slides and then I can go to the learning platform um what is so very often the question is what what are the prerequisites so the prerequisite here is uh the prerequisite here is uh uh Python Programming um and even if you don't know Python Programming we have some tutorials uh that uh we can uh give you um and then you can we can get you up to speed on Python Programming then uh what is the technology stack we use we use uh open Ai and hugging face and llama 2 and symbol AI we recently added this in our last boot camp it's a very special purpose uh uh conversational AI model uh that we um that we uh we talk about uh call them domain specific models so we uh we talk about general purpose models like you know uh open AI GPT series or llama 2 Series and then we also talk about uh domain specific models then uh in Vector databases we have mongodb uh vect and we use chroma in some exercises and redis mainly uh and then uh all of these I mean think of this as Vector databases in nlm cach and then also retrieval augmented generation uh in that space this is our technology stack uh then we uh we have in orchestration we have L chain llama index and then um in in terms of deployment and hosting and compute uh the um ylabs is our partner runpod is a partner ziml is partner so we talk about how to how to do logging how to how to uh how to actually fine-tune the models in a GPU cloud and all of that and deploying the model um as a web app uh we in uh in our exercises we teach people how to actually deploy this app in streamlet so um we start with the fundamentals we make sure that everyone understands the historical evolution of embeddings we don't uh so uh I'm an educator at I mean I love uh uh teaching and one of the things that uh uh one of the philosophies and uh one of the I would say Corner St Stone of our uh our teaching philosophy is that we have to set the context instead of directly going to Transformers hey suddenly I mean what are Transformers what is attention mechanism we actually start with one heart encoding uh of course some people who are already familiar right so they have to bear with us but we start with one heart en coding we explained to people well uh one heart en coding was there then count based was there then uh then there was this tfidf based approaches then there was something called BT uh and then how did all of this evolve and until we landed in uh in in this Transformer based models era so setting the foundations and along the way for everything some of the exercises are there in class and some of them are going to be uh um some of them are going to be actually uh right there in the learning platform you go and click open up and then practice so we will talk about a lot of this in a lot of detail uh as I mentioned all of this uh we will also be talking about Vector databases and Vector databases we'll talk about all the indexing approaches and not just talk about them practical exercises in Python how to create this uh these uh uh how to create these uh um this different kind Vector databases or to create a vector store using some of these indexing strategies then we will also be talking about the retrieval strategies and all the uh optimization and there are some practical exercises similarly uh semantic search um uh we talk about semantic search and then how do you build a semantic search engine using a vector database uh we talk about prompt Engineering in quite a bit of detail we have uh uh we have interesting exercises I will show you of course I cannot show you every single thing that we have uh and given the time that we have but um I will show you some examples of how uh we'll do it uh uh in uh in practice uh we T we spend almost a whole day on Lang chain uh we talk about pretty much uh all the topics uh that are uh relevant in the Contex of Lang chain about 6 to 8 hours uh are spent just on Lang chain so all the document loaders retrievers uh uh chains and memory and so on everything uh you know how do you how do you create agents how do you create a search agent how do you create a Wikipedia agent so we talk about all of that in uh in quite a lot of detail uh let me see there is I mentioned agents as well then we also talk about observ probability and large language model Ops well what is that um you know you have to monitor your model you have to be able to uh you know make sure that your model uh the prompts are not toxic the prompts uh you the responses are not toxic uh are there any guard rails around uh your model or not so we we talk about we spend quite a bit of time on this as well uh we also uh talk about fine-tuning we start with uh fine tuning we talk about quantex Iz ation we talk about Lowa low rank adaptation we talk about QA which is quantized and low rank adaptation based models and we uh build uh we have a practical exercise where uh we teach people how to find you a llama 2 model with which has Lama 2 7 billion parameters and 4bit quantized model and we uh then we run some evaluation exercises compare uh the Baseline model uh Baseline litu model with uh with the with the fine tune model and see what is the relative trade-off between fine-tuning and um U you know an unfin tuned model and fine-tune model and then when we build some rag applications uh we have this practical experience when to use rag versus when to use fine tuning um and then uh we actually uh build a an actual uh application uh you can see uh the list of speakers uh we have one of the uh best speakers out there in Industry all of these people are actually working on building these things right so whatever they teach uh they have practical experience with what they are teaching it's not really I mean I I know this stuff I I I read this paper I um I watched this YouTube to tutorial and I attended this course ER course no I mean these people are practitioners um some of the some of the uh like most uh like I would say most uh experienced people out there in the respective space so that actually makes it a lot of fun uh let me before I will go to our uh I will go to our learning platform I will show you how the entire learning platform is set up um uh you can see that uh we we have our partners some of the leading people in Industry uh we have uh we uh this is going to be the next cohort is going to be our fifth cohort but we are already as I said I mean we are the only company but we are very much respected uh any company that you uh you can see that all of these companies or rather some of the some of the companies have attended we have the logos here um and you can see that we are uh you know already we are getting a lot of fraction from uh from uh industry uh okay so the next boot camp out there is uh let me see the next boot camp is on April 23rd uh you can go to the website maybe the link has already been posted but you can see that the next boot camp is planned for uh April 22nd not 23rd April 22nd to 26 and then um let me actually go first and show you our learning platform uh and then I will uh go into you know the logistics of a few other things so if you look at this uh all the topics that I mentioned you know evolution of text and Bings and attention mechanism and Transformers and Vector databases and semantic search and prompt engineering and finetuning you can see all of this nebula is a domain specific conversational model uh you know Lang chain is there so the way all of this works is let's say we discussed attention mechanism now once the the theoretical discussion is done um we ask people to go and click on well one of these labs and if you can see uh what we have done here is um 40 hours is simply not enough to cover so much content and any time spend in installing the packages any time spent in figuring out dependencies hey my Mac does not have this Library what should I what should I do we have completely eliminated uh that uh that kind of uh uh that kind of um uh you know barrier and what happens is we give people access to uh these our uh um Jupiter notebooks in the cloud um and what people do is everyone gets you see that this is my me here uh my account so everyone gets a dedicated space when you click on this a pod comes up you finish your work you store the work and when you come back you still have this space for you so this compute is it comes with the registration fee that's of course a substantial uh expense on our side but what we want to uh we want it to be a very high quality premium experience so we'll give you the API Keys during the boot camp we give you this uh the boxes for one year uh until you know after the training so you come in you join the training and then you have access to this so anytime when you have to you know practice uh you do not have to worry about hey I mean what should I do um I I don't have this piece of code running I don't have a laptop so it makes it actually reduces the barrier to learning a lot of times the barrier is well I want to learn but I don't know um you know I I'm I'm stuck with this installation so we have removed some of those uh infrastructure or you know inertia that sometimes people people would have because I I don't have um I don't have the right Tools in place um we have other tools like this uh if you look at this this is our prompt engineering sandbox if I go here uh I'm just uh picking up a random prompt here I'm going going to copy and paste this right so now assume the role of soci social media marketer for fashion brand devel developer concise and engaging caption For an upcoming product launch on Instagram I'm going to go and run this and now we will uh we will teach you pretty much all the nuances of prompt engineering and then instead of just hand waving uh hey this is uh you must have a role defined you must have a uh your instructions have to be clear and then you have to persuade and all of that no I mean it's it's not going to be purely um Talk uh will talk and you will come and uh you're going to practice here you can see that number of tokens temperature you can change the temperature uh you know top PE Etc and then uh and you You observe the impact of whatever is happening um firsthand um right in the in your web browser so no software to install at all um I can keep going here you know we have a lot more actually to talk about I think one thing that I should mention is when we when we let's say uh not everything is going to be here we also have other tools in place when we do the Llama 2 fine-tuning exercise uh we ask you to uh we give you uh GPU Cloud your own dedicated GPU Cloud uh credits you get uh credit for finding the Llama 2 model you f tune you rerun and you know just make sure that you understand the underpinnings of everything that is going on okay uh so we have been asked quite a lot I cannot uh attend the in-person training maybe because of budgets maybe because of the time commitment maybe because of travel uh and so on so what we have done is we have uh we are we have announced um smaller chunks of this entire boot camp so think about this boot camp being uh a 40-hour training so if you have broken this down into six or seven trainings and what we will do is um you can come in and um s you can uh you can attend these trainings in an instructor-led online fashion uh so for instance large language models for everyone this is the very barebones in very high level in non-technical or maybe sem my technical introduction to large lmic models uh which is about the first day of our training we talk about that uh then prompt engineering well uh the name says it all what this is going to be about then there is a training on fine tuning there is a training on L chain so what we have done is we have introduced these uh shorter duration trainings that are going to be instructor-led live trainings with all the tools that I showed you just now except that you don't have to travel physically and uh and then also I mean what if you don't want to attend the entire boot camp you're only interested in Lang chain or you're only interested in uh um prompt engineering training and what we will do is uh we will uh will uh offer these as smaller uh instructor Le trainings still the same quality still the same learning but except that you're are not face Toof face in the same room it is going to be uh in the uh it is going to be live uh instructor Le um what else let me actually go and see if I missed anything uh I think that's pretty much it from the presentation side I'm happy to take any questions uh okay so there is this uh question hi Raja and team uh U please correct me I'm I will try to pronounce the name as to the best of my ability car wendan BR uh from India hi Raja in team I'm a data engineer with good experience with python and data science I don't have any experience in AI can I still join the training yes absolutely uh I have been uh I've been teaching uh machine learning for a long time and uh the the funny thing is uh um I have stated this publicly in on social media a while back I mean check out my LinkedIn post right so that 95% of llm application development is still good old software engineering and data engineering because the AI part yes I mean and for the most part any engineer can easily easily wrap their head around it so I absolutely have no doubt if you already are a data engineer uh you will actually see that this is mostly data engineering yes you need to understand Transformers a little bit you need need to understand uh you know a maybe some attention mechanism a bit of Neal networks and we'll will'll do that for you right you do not need to have any background in traditional ml or AI is there a certification available at the end yes we have a certification that is available from one of the oldest universities in the United States um University of New Mexico we are partnering with the University of New Mexico continuing education and we are going to uh well we'll be giving you the um handing you the certificate and you can um uh and you can basically it's a verifiable certificate it's if you are in some professional body that requires you to actually continuously get trained since it is an accredited University you can actually uh use those cus toward some um some kind of uh Professional Credit as well and you can actually request a transcript even though this is more of a uh you know completion right so you can still request a transcript of attending from the University of New Mexico for I think $10 or 10 I I believe it is $115 don't quote me on that later but you get the idea right so it's uh yes we have we are backed by the continuing educ continuing education department at University of New Mexico uh do we need a Mac or uh we can use a PC uh that is exactly what I was telling you uh uh stasti stasti uh this is exactly what I was telling you right so it does not matter right so when you are using these browser based Labs it doesn't matter right so whether you have a Mac or you know you know Ubuntu or a PC it does not matter we are going to you know the labs are going to be launched in the web browser and this is exactly I mean so we I've personally trained more than 8,000 people globally and the company has trained more than 11,000 people right so we have seen so many you so many different challenges in learning and we have removed all those barriers we want to make sure that uh you know people are actually um you know uh people don't spend time the Learners don't spend time in installing libraries and resolving dependencies completely uh platform agnostic as long as you have web browser and some Modern web browser right so a chrome or an edge or a Firefox and that that's it that's all you need uh when is the online boot camp coming please uh check out the website uh the website has the schedule uh if you go on the website data scien dojo.com It Go Under courses and then some of the courses are here and we are adding more courses uh it's uh the courses are ready it's more about our you know our sales and marketing motion to know just uh come into play at the moment but otherwise these are uh already listed on the website okay uh let me see there is there is some questions in the chat let me see yes garindan I've answered your question Sur uh building and deploying via pipelines as well Sur I'm not entirely sure uh but uh what we do here is I can give you an idea building and deploying uh we pipeline L yes we use the state-of-the-art tools for uh um for we are I mean this is a no fluff boot camp right so people are generally they are quite tired by the end of the boot camp right so it is some serious work um we have multiple instructors there uh and but this is some serious work right so you have some work to do uh and then we make sure that you do this work uh one more thing is sometimes people actually may get a bit discouraged hey I don't know coding notice how how this effective this approach is when you click on this we have had people who had minimal to no coding background but after the discussion uh after the discussion we had now notice this right so all you have to do in a Jupiter notebook let me pick an easier Jupiter notebook just to make a point here so all you have to do is run run and run and run and keep going so if you are a technical product manager if you are a Founder if you are someone who does not need to actually write code but still you want to understand it deeply enough that you can guide your team or you can guide uh you know the pro the direction of the product I actually strongly feel we have been we have a product uh and we we are offering some Services as well I strongly feel that um anyone um anyone who wants to be a product manager in llm applications Bas they must understand all of this right so there are so many architectural tradeoffs that you have to make uh so many uh cultural trade-offs too many adoption related trade-offs you know um should we fine tune or should we do rag uh should we use a 4K model um or a 16k model or a you know 128k model um and when to use which one um we have this uh we have built a tool uh internally uh and then uh you know we have these engaging discussions and what I've come to this conclusion right uh that uh you you you have to be uh if you even if you are a program manager project manager or a product manager or a Founder in this space you have to understand um uh understand some of these trade-offs otherwise it is going to be hard for you to actually build something that people need let me see uh um given how quickly technology is evolving this is a question from sures uh given how quickly the techn technology is evolving and Tech uh is becoming old if not obsolete how much of this course is insulated from that reality it is not insulated from that reality we are constantly uh keeping up constantly catching up so I I I think when we started let let me tell you this right so in our first offering of the boot camp we did not talk about um if I remember correctly we did not talk about semantic caching uh we did not talk about uh domain specific models um we we did not talk about hybrid search at that time U there are uh uh you know there are there are so many things so in just the rag space um it is hard to keep up even for me I try to keep up but it is hard for us to keep up but we have a big team of Engineers uh who are also contributors to the curriculum and Labs that we create so as a result of that anything new that they learn they actually would incorporate that in uh in the curriculum let me actually go through this right so this uh this thing uh nebula nebula was not there uh domain specific conversational model it was not there in the first uh offering of the boot camp um guard rails was there minimal we added it later uh evaluation was there very minimally we actually enhanced the model uh enh enhance that module uh substantially um and then uh let me see I hope you get the idea and lot of you know we we were discussing Rag and Vector databases but there are a lot of uh nuances within that um that uh we uh we we did not uh start with it but we have been constantly up uh updating Rahul what is the expected outcome after this boot camp um uh so so uh Rahul so you have given this example does it make non AI Engineers into AI Engineers or non- AI managers into AI managers five five days seems short right so so Rahul um I'm known to be a like a straight shooter to the point sometimes people don't like my answers right so uh we are not in the business of making you know um selling snake oil here right so really the outcome depends upon uh you know who you are what your background is you know I cannot promise any outcome based on without knowing you right but go and check out uh I will bring up our uh website here go and check out I mean we have some of the very seasoned industry leaders uh let me go to reviews there are people hundreds of people on camera more than any other boot camp in Industry uh who are actually talking about it go check out our testimonials video testimonials case studies these people are all I mean there is a NeverEnding list of people who have attended our trainings and they think we are the best train best decision that they have made ever right so go check out I I don't know if you are interested knowing uh you know Microsoft just go and search I mean there are some people who are like really seasoned people in Industry who are working for some of the top companies real people and I can I can confidently tell you go and talk to any of them these are maybe a few thousand people who have attended and they and they will tell you you know what they think right so David for instance he's right here uh David actually attended our data science boot camp and U he actually according to him thinks fundamentally changed for his company and now he came back to attend our uh our llm boot camp all the way from Paris right so so you know we are Erica also she's a returning customer um and then uh Fahad flew all the way from Saudi Arabia so you can go and see Bernard a repeat customer uh so just go and check it out right so what I would I can tell you is that if anyone can teach it well we are the ones but whether you will get I mean you will turn something non- AI to AI I am very confident that it should happen but uh five days seems short and once you start attending you will actually realize that it is not it is very very intense you you will be you will be busy throughout the week um will there be a project experience even for online boot camp uh project experience we are trying to figure out because the courses are actually now um you know in the in-person experience we are actually talking about you know one boot camp that covers everything that we'll be covering online but when you talk about uh when you talk about the online experience the problem is that uh we cannot have a 5 day eight hours a day 9 hours a day training it's uh it's that kind of uh uh engagement it's not going to be there in online so we have decided to break it down but when you break it down you know where do you do the project we can certainly uh Caron if you can maybe you should set up a time uh someone I think shared from the team someone shared a link set of a time I'm happy to talk to you right so I would like to understand how I can help right so happy to have it talk to you uh then uh uh baset I have questions uh that llms are computationally heavy models if we provide llm based solution service to the Enterprise and they want a custom model how to convince them on computation is there any reliable solution baset I wish I could answer this in the limited amount of time that is there um and then um because without knowing what the Enterprise is uh um I I can't say actually I honestly uh don't know whether uh how to actually answer your question because a lot of these things are very situational right so it is never one size fits all uh in some cases and I can give you an example for one of the customers that we were talking to um you know some customers would say we are very price conscious others would say no I don't care right so give me a solution uh for and uh not exaggerating or making it up right so one customer I talked to they said if you can give me the correct answer every time right now there is a human actually doing this for me uh think about this some research and uh compilation work if you can build this for me I'm even if a single query takes $350 up to $350 a single query a single summarization because this is what the cost is on their side actually the cost was around7 $800 so they said if an llm can give me uh this much uh so you know sometimes you would and some customers they would say hey I mean I don't want my cloud bills to be so much so really I cannot answer this question and these are the kind of anecdotes these are the kind of practical insights these are the kind of discussions that we have during the uh during this uh boot camp in person or during the training let me see there are more questions um do you have plans to launch a full online course anytime in the future traveling traveling and staying will be expensive for out of city folks I completely understand and this is why we have this and then um um and but at the same time uh you know we also have to consider uh the product Market fit uh uh question right so how many people would be willing to attend the same boot camp for five straight days you know okay Rahul you would right so I love that right I love the spirit of this uh but Rahul what if we broke uh we have we broke that uh broke the entire boot camp down maybe you can take them over five weeks so what we will do is we'll rotate so think about this that the boot camp has certain uh the boot camp has certain uh you know coherent modules one is let's say prompt engineering one is uh uh Lang chain one is fine-tuning one is guard rails and deployment one is uh building a rag uh you know more of the nuances of rag and one is Transformers and detention mechanism so what if we broke them down and offered them perhaps you know uh right now the the way it is done is that you have a u um a dur um one hour or rather one day you finish one of these then next week one day you finish one of these so you can still actually combine all of them you can still combine them uh so but doing all of this in a single week I think uh our judgment call is that it is going to be hard and of course I mean we have uh we have to make sure that uh there is a sizable audience but if uh for those of you are attending I was told that we have 300 people in audience today but if you are attending uh if you P trending and if you think that you would like to do a 5day online version I'm happy to do it right so I enjoy teaching I I love it so just drop us a note so and then if you have enough subst enough people who would sign up then we we can announce it I mean that should not be a problem it is just our judgment call that we are doing it you know as separate courses as opposed to a single course okay um uh once we finish um the boot camp and if more update are done to the boot camp curriculum will we get some discount to learn those up actually uh yes uh you don't even have to pay don't worry about discounts any new curriculum anything that is added I mentioned this that you have access to this learning platform for one year so anything that we do for instance Lewis Sano was not actually um when we started Lewis was not there uh but later on uh you can see that Lewis taught for the first time Transformers and attention mechanism and you can see that uh uh when we uh brought him in the video was actually uploaded and uh so any uh any changes in the curriculum uh any changes in the curriculum any new Labs that are added you will have access to them for sure uh then is there a onetoone mapping between each section of inperson boot camp to online for the most part yes um yes uh you know it is uh it is really taken from the it is really taken from the existing uh inperson boot camp curriculum and then uh then mapped it to uh you know to break it down into different components and uh Stasi uh uh and do we get any help in getting jobs no we are not a job placement boot camp at the moment uh even if you are in the USA however uh we have helped people but this is not uh something something that we promise as part of the product this is something here we get to know you uh and then um we you know we have a fairly big Network and we may if someone is looking and if you're the right fit you may people have found jobs but this is not something that we promise fundamentally I hope that makes sense uh then we have uh any chance you would have a similar boot camp in NYC we have been planning uh for NYC uh but nothing uh in the next two to three months mon um I would attend this boot camp online if done in a whole week uh but right now it is impossible for me to go to Seattle now um yes sasti we would love to help uh in any way we can uh and then as soon as we have uh you know as soon as we are um we are up to speed on you know and we rather we have enough people in this uh in a in a 5day setting we are going to be uh we will let you know okay um let me see [Music] therefore okay I think that should be it unless there are any other questions I will uh probably you to call it a day okay one more thing are the online courses sequenced I see that M for everyone scheduled for April but Lang chain is on March 26th will it be possible to sequence it in the same way it is done um uh yes you can I mean uh these are going to be repeat offerings uh so you can actually sequence them you can sign up for them and uh create a sequence on your own uh for sure but we have to you know come up with a Cadence um um in terms of you know we have our own logic Logistics as well uh but uh for the most part these are going to be sequenced once we streamline the process uh they are going to be SE okay sounds good then uh I think uh since there are no more questions I will I will end this call thank you so much everyone and I'm looking forward to seeing you some of you at the boot camp have a great rest of the day